CLASSIFICATION OF IMBALANCED REMOTE-SENSING DATA BY NEURAL NETWORKS

The multilayer perceptron neural network has proved to be a very effec tive tool for the classification of remote-sensing images. Unfortunate ly, the training of such a classifier by using data with very differen t a priori class probabilities (imbalanced data) is very slow. This pa per describes a learning technique aimed at speeding up the training o f a multilayer perceptron when applied to imbalanced data. The results obtained on an optical remote-sensing data set suggest that not only is the proposed technique effective in terms of training speed but it also allows classification results to be more stable with respect to i nitial weights. (C) 1997 Elsevier Science B.V.